For ligation of a linear molecule to occur, the two ends must come together at the active site of the DNA ligase. In a simple molecular cloning experiment the aim is usually to avoid recircularisation of a cut plasmid vector, and instead to get a new fragment of DNA inserted into the vector. However because the ends of the linearised circular plasmid are tethered together (by being ends of the same molecule) they are more likely to encounter each other than they are to encounter the end of an incoming fragment. This is usually overcome, in part, by increasing the relative concentration of the target fragment.
If you are aiming to recircularise a long fragment then the opposite effect will come into play: the ends of individual long molecules will be much more independent and so the end of another molecule will be likely to compete successfully for ligation. In theory this effect could be overcome by reducing the DNA concentration, but I assume that this is impractical for other reasons.